The purpose of this application is to explore how high level planning signals can be used for control of neural prosthetics to assist paralyzed patients. This study has both scientific and engineering components. The scientific investigations entail exploring how cognitive signals related to movement intentions are encoded in the parietal-frontal circuits. The engineering component will be guided by the scientific findings to best design and tailor algorithms for decoding these cognitive signals. Areas of algorithmic development include new signal processing and feature extraction techniques, extensions of Bayesian classification and Kalman filtering algorithms, and new applications of speech recognition and finite state machine techniques. Aim 1 will examine how the goals of reach movements are represented in 3 dimensions in the parietal reach region (PRR) and the dorsal premotor cortex (PMd) and develop decode algorithms to control the location of a cursor using these signals (so-called brain-control task). Goal decoding has the attributes of being very versatile and rapid for prosthetics applications. This aim will also determine if goal locations can be decoded using local field potentials (LFPs) rather than spike activity using advanced signal processing techniques. An advantage of LFPs for prosthetics is their ease and longevity of recording. Aim 2 will examine whether neural activity in PRR and PMd predicts the current location of the limb during trajectory movements, and if this "forward model" can be used to generate trajectories in brain control tasks. Techniques suited for continuously varying dynamic systems will be applied to decoding the trajectories. Aim 3 will study plasticity in PRR and PMd related to context, learning and reward. In this aim we will examine how the ability of the brain to learn and adapt can lead to better performance of brain-machine interfaces. Aim 4 will examine the very challenging situation of decoding movement plans continuously. Studies in this field generally use event markers derived from the trials of a task to assist decoding. However, these markers will not exist for clinical applications of prosthetics and the problem of recognizing and interpreting neural signals becomes much more challenging. We will apply and extend techniques from speech recognition and finite state machines to this problem. In particular, we will examine how eye movement information during natural hand-eye coordination can help decode reach movements from neural activity. This eye movement information will be derived from eye movement recordings and from the recording of neural signals related to eye movements. Knowledge from this work will be applied to brain- control tasks involving the continuous, sequential determination of goals. [unreadable] [unreadable] [unreadable]